I recently accepted a new job. From October 2009 on, I'll be a junior researcher at the CWI (Centrum Wiskunde en Informatica), a national research center here in Amsterdam. I'm very happy to be working with Prof. Han La Poutre.
I will actually be a Phd student but do research in a predefined context - nowadays, a lot of Phd positions in the exact sciences are a mixture of research work and research exploration. An exciting part about this project is that it is about a very real challenge and there are many stakeholders involved whose views on the problem are important. I will do fundamental research on dynamic and adaptive multi-agent systems, but I will also be busy wearing the hats of different stakeholders and try to bring their view into the model.
[image via http://www.usgbc-centraltexas.org]
The project I'll be working on is called IDeaNed (Intelligent en Decentraal Management van Netwerken en Data, only dutch description so far). It deals with one of the big challenges we face in the next years - a redesign of our energy distribution networks.
As we begin to exit the age of fossil energy, energy will soon become a problem for us, not a cheap and abundant catalysator for progress. Now, as an AI researcher, I can't help out where it is most important - say, actually invent an effective renewable energy source or useful, longlasting batteries. But what we need for sure is a new energy infrastructure. There is already a term for this - the "Smart Grid"*.
Our existing energy networks are of an old, centralised design which is not suitable for the way we need to look at interconnected things. As the report "Grid 2.0" (PDF) puts it:
"It is hard to make the link between flicking a switch and the distant power station that made it possible to turn the light on.
Partly as a result of this lack of a feedback mechanism, and partly because of technological constraints, Grid 1.0 is surprisingly inefficient. Only around 40 per cent of primary energy input (coal or gas) used in power stations is converted into usable electricity, the rest is wasted heat. A further nine per cent is lost as the power moves through the transmission and distribution system. Then a further third is lost in our homes and offices because they are poorly insulated, not designed with energy in mind, and inhabited by people who do not see themselves as players in the energy game."
We need a network that is decentralised (can work with input from several local sources, not only big central plants, can communicate locally about sharing loads) and avoids efficieny problems in peak times. Such a net needs real-time price mechanisms and needs to accomodate a lot of players: customers want to use effectively, producers want to sell efficiently, governments want to distribute evenly **. Meanwhile, there is a real, non-abstract component: the network. Voltage capacities at different parts of different sub-networks dictate what is possible and what is expensive to do. A lot of these things change all the time. This can get very complex very fast.
What we'll be doing is modeling Multi-Agent systems in order to learn what good price mechanisms are and what good automisationable strategies of the local players could be. We will get input from another Phd student from Eindhoven supplying technical findings about the infrastructure and we'll work with companies in the dutch energy market (big players and consultancy agencies that develop energy network simulation software). This highly integrated approach of designing scientific projects is a unique dutch approach and I am quite excited about it.
This topic is not only being picked up in the Netherlands. Big companies like GE or IBM are already promoting their competence in this topic which isn't even entirely understood yet.
They smell money in creating a market better targeted at our actual energy usage and in savings due to efficiency. This might be one of the rare cases where what they want and what society needs have a significant overlap. For instance, I recently learned that energy efficiency on the last mile is multiple times more effective than energy efficiency at creation time (due to all the energy losses by conversion or distance that already accumulate until the last mile). And it is easier to do.
* There might even emerge a bubble with all the money being thrown at this topic right now by governments, but I hope that in 4.5 years, when I finish this project, things will have come to terms, the crooks took their money and left or were thrown out and we know where to go.
** Especially when there is not enough. In a post-fossil world there may be "bad weeks" with few energy available. Or, in more grim scenarios, we'll have few energy at all times and almost none in "bad weeks". Making sure distribution can be fair is crucial and not as easy as it first sounds.